.Failure time regression to determine warranty policies of computers

Abstract

The present era of consumerism has substantially redefined the relationship of buyer and seller. Heightened consumer expectations have placed more responsibility on the manufacturer for the performance of his goods and a warranty or a written affirmation of the quality or performance of the product is a requirement in most cases. However, little attention has been given by the manufacturers for the scientific setting of warranty policies on computers. Thus this paper explores a method to estimate warranty periods based on failure time of computers. The data pool consists of two data sets, namely repair data and sales data from a reputed IT solution company. For each computer, the installation date is taken from sales data and failure data is taken from repair data and the failure time is calculated. A weibull model was found to fit the data well. Cumulative probabilities of failures based on this model were taken in to account in calculating suitable warranty periods. Graphical goodness of fit consisting of probability plots and simultaneous equal precision (EP) confidence bands are used to assess the validity of distributional assumptions. The main problem encountered in this study is setting suitable warranty periods. Life times of the computers were used as the basic modelling variable. The process of solving the problem consists of various statistical methods and techniques. Since there are various engineering methods to calculate warranty periods, these statistical methods show a different angle. To illustrate a valid warranty period certain distributional assumptions were used and those assumptions laid a base for solving the problem